Skip to content
intermediateaudio

Continuous Emotion Rating

Rate emotional dimensions (valence, arousal, dominance) continuously following MSP-IMPROV protocol.

Get This Design

This design is available in our showcase. Copy the configuration below to get started.

Quick start:

bash
# Create your project folder
mkdir emotion-dimensional-rating
cd emotion-dimensional-rating
# Save config.yaml from this page into this folder
potato start config.yaml
Continuous Emotion Rating annotation interface preview

Configuration File

yaml
annotation_task_name: "Continuous Emotion Rating"
task_dir: "."

port: 8000

# Data configuration
data_files:
  - "data/speech_utterances.json"

item_properties:
  id_key: id
  text_key: text
instance_display:
  fields:
    - key: audio_url
      type: audio
      label: "Utterance"
    - key: text
      type: text
      label: "Transcript"

# Annotation schemes
annotation_schemes:
  # Valence (pleasure dimension)
  - annotation_type: slider
    name: valence
    description: "Valence: rate the pleasantness of the emotional state. 1 is very negative (unhappy, annoyed, dissatisfied), 7 is very positive (happy, pleased, satisfied)."
    min_value: 1
    max_value: 7
    step: 0.1
    starting_value: 4

  # Arousal (activation dimension)
  - annotation_type: slider
    name: arousal
    description: "Arousal: rate the activation/energy level. 1 is very calm (relaxed, sleepy, bored), 7 is very activated (excited, alert, energetic)."
    min_value: 1
    max_value: 7
    step: 0.1
    starting_value: 4

  # Dominance (control dimension)
  - annotation_type: slider
    name: dominance
    description: "Dominance: rate the perceived control/power. 1 is submissive (controlled, influenced), 7 is dominant (controlling, influential)."
    min_value: 1
    max_value: 7
    step: 0.1
    starting_value: 4

  # Naturalness
  - annotation_type: likert
    name: naturalness
    description: "How natural does the emotional expression sound?"
    size: 5
    min_label: "Very artificial"
    max_label: "Very natural"

  # Emotion clarity
  - annotation_type: likert
    name: clarity
    description: "How clear/recognizable is the emotional state?"
    size: 5
    min_label: "Ambiguous"
    max_label: "Very clear"

  # Primary emotion (optional categorical backup)
  - annotation_type: radio
    name: primary_emotion
    description: "If you had to pick one emotion label, which fits best?"
    labels:
      - Anger
      - Happiness
      - Sadness
      - Neutral
      - Fear
      - Surprise
      - Contempt
      - Mixed/unclear
    required: false

# User settings
require_password: false

# Output
output_annotation_dir: "annotation_output/"
export_annotation_format: "jsonl"

Sample Data

json
[
  {
    "id": "emotion_dimensional_rating_1",
    "audio_url": "https://upload.wikimedia.org/wikipedia/commons/5/58/En-us-I%27m_in_love_with_you.oga",
    "text": "I'm in love with you."
  },
  {
    "id": "emotion_dimensional_rating_2",
    "audio_url": "https://upload.wikimedia.org/wikipedia/commons/9/9d/En-us-I_know_you_are_but_what_am_I.ogg",
    "text": "I know you are, but what am I?"
  }
]

// ... and 3 more items

Details

Annotation Types

sliderlikert

Domain

AudioSpeech

Use Cases

emotion ratingdimensional emotionaffective computing

Tags

audioemotiondimensionalvalencearousal

Related Designs

Audio Transcription Review

Review and correct automatic speech recognition transcriptions with waveform visualization.

likertmultiselect

Audio-Visual Sentiment Analysis

Rate sentiment in speech segments following CMU-MOSI and CMU-MOSEI multimodal annotation protocols.

likertradio

LibriSpeech - Audio Transcription and Quality Rating

Audio transcription quality assessment based on the LibriSpeech corpus (Panayotov et al., ICASSP 2015). Annotators rate audio quality using a slider and classify audio segments by content type, supporting speech recognition research and dataset curation.

slideraudio_annotation